Thermodynamic Studies of the Mechanism of Metal Binding to the Escherichia coli Zinc Transporter YiiP*

Sequence homology of the Escherichia coli YiiP places it within the family of cation diffusion facilitators, a family of membrane transporters that play a central role in regulating cellular zinc homeostasis. Here we describe the first thermodynamic and mechanistic studies of metal binding to a cation diffusion facilitator. Isothermal titration calorimetric analyses of the purified YiiP and binding competitions among Zn 2 (cid:1) , Cd 2 (cid:1) , and Hg 2 (cid:1) revealed a mutually competitive binding site common to three metal ions and a set of noncompetitive binding sites, including one Cd 2 (cid:1) site, one Hg 2 (cid:1) site, and at least one Zn 2 (cid:1) site, to which the binding of Zn 2 (cid:1) exhibited partial inhibitions of both Cd 2 (cid:1) and Hg 2 (cid:1) bindings. Lowering the pH from 7.0 to 5.5 inhibited binding of Zn 2 (cid:1) and Cd 2 (cid:1) to the common site. Further, the enthalpy change of the Cd 2 (cid:1) binding to the common site was found to be related linearly to the ionization enthalpy of the pH buffer with a slope corresponding to the release of 1.23 H (cid:1) for each Cd 2 (cid:1) binding. These H (cid:1) effects are consistent with a coupled deprotonation process upon binding of Zn 2 (cid:1) and Cd 2 (cid:1) . Modification of histidine residues by diethyl pyrocarbonate specifically inhibited Zn 2 (cid:1)

Zinc is required by many metalloproteins for catalytic activities, structural stability, and functional regulations (1)(2)(3). Low cellular zinc levels inhibit cell growth and division, whereas high zinc levels are toxic. To maintain cellular zinc content within a narrow physiological range, cells rely on a complement of zinc homeostatic mechanisms by the expression of zinc chelation proteins, sequestrations of zinc into intracellular membrane compartments, and controls of zinc entry and exit from cells through several families of zinc transporters (4 -9). Among them, zinc transporters in the CDF 1 family play a major role during zinc excesses, conferring tolerance for zinc and some divalent transition metal ions (10). Seven mammalian CDF proteins, ZnT1-7 (zinc transporter 1-7), have been cloned and characterized. ZnT1 is involved in zinc efflux across the plasma membrane (11), and ZnT2-7 facilitate zinc accumulation in various intracellular compartments (12,13). Bacterial CDF proteins are distantly related to mammalian counterparts by a homologous hydrophobic domain with six distinct hydrophobic segments (14,15). Despite large variability in the hydrophilic regions, all CDF family members identified so far, prokaryotic or eukaryotic, appear to transport metal ions exclusively (10). The migration of a metal ion in a transporter involves temporary association of the metal ion with one or more binding sites along a translocation pathway. This combined process of equilibrium binding and energized movement brings about the metal selectivity of the transporter and the mobility of the metal substrate. Therefore, thermodynamic and kinetic studies of CDF proteins are critically important to the understanding of the molecular mechanisms of metal ion transport across biological membranes.
The kinetics of a homologous Escherichia coli CDF protein, ZitB, was studied by stopped-flow measurements of metal ion fluxes across the membrane of proteoliposomes reconstituted from the purified ZitB (16). The ZitB-mediated transport was shown to be a substrate-saturable process that can be described by a two-step reaction of an equilibrium binding followed by a conformational transition that moves the bound metal ion across the membrane. Both Cd 2ϩ and Zn 2ϩ are effective substrates, the translocation of which requires a proton movement in the reverse direction of the metal transport. This coupling of metal ions to protons provides the energetic basis for zinc efflux in E. coli, where the downhill influx of proton produces free energy to drive efflux of the cytoplasmic zinc.
To gain an energetic and mechanistic understanding of the metal ion binding to CDF proteins, we used isothermal titration calorimetric (ITC) to directly measure the heat exchanges that accompany metal binding to an E. coli CDF protein, YiiP. This calorimetric approach allows the dissecting of the Gibbs free energy of binding ⌬G into two thermodynamic components: the enthalpy change, ⌬H, and the entropy change, ⌬S (17). ⌬H is measured directly from ITC measurements; the binding constant K a and stoichiometry are obtained by a leastsquares fit of the binding isotherm to a binding model. ⌬G is given by ⌬G ϭ ϪRTlnK a , and ⌬S is calculated by using the standard thermodynamic expression ⌬G ϭ ⌬H Ϫ T⌬S. These thermodynamic parameters permit the evaluation of enthalpic and entropic contributions to the Gibbs free energy of binding, thereby providing a thermodynamic description of the binding reaction. Furthermore, the coupling between metal ions and protons can be evaluated by the dependence of the metal ion binding on pH and the dependence of the apparent binding enthalpy (⌬H app ) on the proton ionization enthalpy (⌬H ion ) of the buffer in which the binding reaction takes place (18). In the latter case, ⌬H app is the sum of two enthalpic contributions, the metal ion binding enthalpy, ⌬H bind , and the proton ionization enthalpy of the buffer, which gives rise to a cumulative relationship, ⌬H app ϭ ⌬H bind ϩ n⌬H ion , where n is the number of protons that are absorbed or released upon metal binding (19).
Here we describe the thermodynamic analyses of metal ion bindings to the purified YiiP. YiiP was over-expressed, solubilized by detergent, and purified free of metal contaminants prior to ITC experiments. Titrations of YiiP were carried out using three group-12 metal ions: Zn 2ϩ , Cd 2ϩ , and Hg 2ϩ . Zn 2ϩ is a borderline soft-hard metal ion with a highly concentrated charge and strong electron affinity, whereas Hg 2ϩ is a soft metal ion with a bulky and highly polarizable charge. Cd 2ϩ , known as a substitute for Zn 2ϩ in many zinc enzymes, has an intermediate softness (20). Chemical differences in the Zn 2ϩ 3 Cd 2ϩ 3 Hg 2ϩ series resulted in distinctive calorimetric responses, serving as thermodynamic signatures to aid the deconvolution of complex binding isotherms. Our results provide mechanistic insights into the molecular recognition that underlies metal ion binding to a metal transporter.

EXPERIMENTAL PROCEDURES
Cloning and Expression Plasmid Construct-The entire open reading frame sequence of YiiP was obtained by PCR using the genomic DNA of E. coli BL21 strain as template and a pair of YiiP-specific primers with a NdeI and a BamHI site incorporated into the 5Ј-ends of the forward and reverse primer, respectively (forward primer 5Ј-GCAGCCATAT-GCTCGAGAATCAATC-3Ј, reverse primer 5Ј-GGATCCTTATGAAAG-CATAGACCGT-3Ј). The PCR product was double digested using MfeI and BamHI (New England BioLabs, Beverly, MA), yielding two fragments. The larger piece was isolated and further digested using NdeI. The resulting NdeI-MfeI and MfeI-BamHI fragments were inserted between the NdeI and BamHI sites in expression vector pET15b (Novagen Inc., Madison, WI) in frame with an N-terminal 6-histidine affinity tag followed by a thrombin proteolytic cleavage site. The completed expression construct pHis-TB-YiiP was verified by automatic sequencing of both strands.
Over-expression of YiiP-Over-expression of YiiP was achieved using the overnight express autoinduction systems-1 (Novagen) based on an auto-inducing procedure 2 (22). BL21(DE3)pLysS cells (Novagen) were transformed with pHis-TB-YiiP and stored in a noninducing medium at Ϫ80°C. The frozen stock was inoculated into 50 ml of Luria-Bertani broth containing 100 g/ml ampicillin, and the culture was grown to log phase at 37°C, whereupon cells were harvested and then inoculated to a 6-liter auto-induction medium for unattended induction and expression of YiiP. The overnight auto-induced culture typically reached an optical density (A 600 ) of 4-6 absorbance units.
Purification-Cells were harvested and lysed by three passages through an ice-chilled microfluidizer (Microfluidics, Newton, MA) at 1000 p.s.i., and the resulting membrane vesicles were collected by centrifugation at 140,000 ϫ g for 45 min. The pellet was solubilized using a detergent buffer (100 mM NaCl, 20 mM HEPES, pH 7.5, 7% n-dodecyl-␤-D-maltopyranoside (DDM; Anatrace, Maumee, OH), 0.25 mM TCEP, 20% w/v glycerol) with gentle stirring for 40 min at 4°C to achieve a complete extraction of YiiP from membrane vesicles. Insoluble materials were pelleted by centrifugation (140,000 ϫ g for 45 min), and the supernatant was passed through a Ni 2ϩ -nitrilotriacetic acid superflow column (Qiagen, Valencia, CA) at a flow rate of 2 ml/min. After washing the column with 10 bed volumes of wash buffer (20 mM HEPES, pH 7.5, 300 mM NaCl, 20% w/v glycerol, 0.05% DDM, 0.25 mM TCEP, 30 mM imidazole), YiiP was eluted with an elevated imidazole concentration at 500 mM. The purified His-YiiP was loaded to a 10-kDa cutoff dialysis cassettes (Pierce) for simultaneous dialysis and protein concentration using an external dialysis buffer (20 mM HEPES, pH 8.0, 100 mM NaCl, 20% w/v glycerol, 0.05% DDM, 0.25 mM TCEP) with the addition of 3% polyethylene glycol 35,000. Thrombin (Novagen, Madison, WI) was added into the dialysis cassette at a ratio of 1 unit/mg of His-YiiP to cleave the N-terminal poly-His peptide.
Preparation of Protein Sample for ITC Measurements-Protein aggregates and trace amount of metal contaminants were removed by size exclusion HPLC prior to calorimetric titrations with metal ions. Concentrated YiiP, typically at a concentration of 10 mg/ml, was incubated with 5 mM EDTA for about 20 min before being loaded to a TSK 3000SW XL column (TosoHaas, Montgomeryville, PA) pre-equilibrated with a degassed mobile phase (20 mM HEPES, pH 7.0, 100 mM NaCl, 12.5% glycerol, 0.05% DDM, 0.25 mM TCEP). HPLC purification was performed using a System-Gold HPLC system run at a flow rate of 0.5 ml/min (Beckman Coulter, Fullerton, CA). TCEP was omitted from the mobile phase when YiiP sample was prepared for Hg 2ϩ titrations because the presence of this reducing reagent caused significant dilution heats of Hg 2ϩ . The HPLC-purified YiiP was detected by its UV absorption at 280 nm, collected as a discrete peak fraction, and then immediately used for ITC measurements. For titrations at low pH, YiiP was purified in an otherwise identical HPLC mobile phase except that 20 mM MES was used to buffer pH at 5.5. For titrations at pH 7.0 in the presence of a different pH buffer, YiiP was purified as described above, except that HEPES was replaced by an indicated buffer. The concentration of the purified YiiP was determined by BCA protein assay (Pierce).
Isothermal Titration Calorimetry-ITC measurements were carried out on a MicroCal MCS titration calorimeter (MicroCal, Inc., Northampton, MA). Metal titrants (chloride salts of Zn 2ϩ , Cd 2ϩ , or Hg 2ϩ ) were dissolved in the same HPLC mobile phase in which YiiP was purified, and their concentrations were adjusted to be 30 -50-fold higher than that of YiiP, typically in the concentration range of 1 to 3 mM. The titrant and YiiP sample were thoroughly degassed before each titration. YiiP samples ranging in concentrations from 0.032 to 0.090 mM were placed in a 1.4-ml reaction cell, and the reference cell was filled with deionized water. All titrations were performed at 25°C. After temperature equilibration, successive injections of an indicated titrant was made into the reaction cell in 5-l increments at 210 -360-s intervals with stirring at 300 rpm to ensure a complete equilibration. The resulting heats of reaction were measured over 30 -50 consecutive injections. Control experiments to determine the heats of titrant dilution were carried out by making identical injections in the absence of YiiP. The net reaction heat was obtained by subtracting the heat of dilution from the corresponding total heat of reaction. The titration data were deconvoluted based on a binding model containing either one or two sets of noninteracting binding sites by a nonlinear least-squares algorithm using the MicroCal Origin software. The binding enthalpy change ⌬H, association constant K a , and the binding stoichiometry n were permitted to float during the least-squares minimization process and taken as the best-fit values.
Reaction with DEPC-YiiP was mixed with 10 mM DEPC dissolved in the HPLC mobile phase with the addition of 2 mM Zn 2ϩ , Cd 2ϩ , or EDTA as indicated. After incubation with DEPC for an indicated time, the reactions were terminated by adding 20 mM imidazole to quench the unreacted DEPC. Next, 5 mM EDTA was added to chelate metal ions in the reaction mixture, resulting in metal-free and DEPC-modified YiiP that was further purified by size exclusion HPLC as described above. The protein peak fraction was collected and immediately loaded to the MicroCal MCS titration calorimeter for ITC measurements.

RESULTS
Purification-The over-expressed His-YiiP was purified to homogeneity by a one-step nickel affinity chromatography (Fig.  1A). His tag and residual metal contaminants in the purified His-YiiP sample are two major factors that interfere with metal calorimetric titrations. Therefore, the His tag was cleaved by thrombin digestion, and the completeness of the proteolytic cleavage was verified by matrix-assisted laser desorption/ionization time-of-flight (MALDI-TOF) mass spectrometry and by Western blot analysis using an antibody specific to poly-His peptide (data not shown). Residual labile metal ions in the YiiP sample were chelated by incubation with 5 mM EDTA for 20 min, but prolonged EDTA incubation caused irreversible YiiP aggregation. The cleaved His tag peptide, metal ion, and EDTA contaminants were removed from YiiP samples by preparative size exclusion HPLC purification. The quality of the resulting YiiP was assessed by analytical size exclusion HPLC, showing a major mono-disperse species with a retention time corresponding to an apparent molecular mass of 190 kDa (Fig. 1B).
Calorimetric Titrations of YiiP with Zn 2ϩ -The energetics of Zn 2ϩ binding to YiiP was examined directly by ITC at 25°C, pH 7.0, as described under "Experimental Procedures." Examples of heat changes accompanying the binding of incremental additions of Zn 2ϩ to YiiP are shown in the upper panels of Fig. 2. Plots of the integrated heat per mole of Zn 2ϩ as a function of the molar ratio of Zn 2ϩ to YiiP are displayed in the lower panels. As shown in Fig. 2A, Zn 2ϩ titrations began with an exothermic heat reaction, which was followed by a late endothermic reaction. This characteristic exothermic-to-endothermic transition suggests the presence of at least two sets of independent Zn 2ϩ binding sites, accounting for the exothermic and endothermic heat reactions, respectively. Furthermore, the dual heat reactions were reduced to a largely endothermic reaction when YiiP was titrated with Zn 2ϩ in the presence of either 0.5 mM CdCl 2 (Fig. 2B) or 0.25 mM HgCl 2 (Fig. 2C). This conspicuous loss of exothermic heat profile in the presence of Cd 2ϩ or Hg 2ϩ suggests that competitive binding occurs at the Zn 2ϩ exothermic binding site (Zn 2ϩ site 1) by Cd 2ϩ or Hg 2ϩ . In contrast, binding to the Zn 2ϩ endothermic binding site (Zn 2ϩ site 2) was less affected by Cd 2ϩ and Hg 2ϩ . Based on this qualitative assessment, Zn 2ϩ binding isotherm was fitted to a binding model containing two sets of independent binding sites, with ⌬H, K a , and binding stoichiometry n values permitted to float. The fit to this binding model yielded 1.5 and 0.84 equivalents of Zn 2ϩ bound to site 1 and site 2 with respective k a values of 0.33 Ϯ 0.47 M Ϫ1 and 6.3 Ϯ 6.1 mM Ϫ1 . The data were of insufficient quality to determine the Zn 2ϩ binding parameters with certainty (Table I), in part because of an unresolved enthalpic transition that was evident at the beginning of Zn 2ϩ titrations, where the exothermic heat effects increased progressively rather than staying at a steady level ( Fig. 2A). This additional heat reaction might be associated with a high affinity binding site on the interface of the YiiP oligomeric complex, because the corresponding transition in the binding isotherm occurred at a stoichiometric equivalence point with a value far less than unity.
Calorimetric Titrations of YiiP with Cd 2ϩ -In contrast to the mix of endothermic and exothermic heat reactions during Zn 2ϩ titrations, the heat effects generated by the Cd 2ϩ binding were purely exothermic with a transition occurring at a stoichiometric equivalence point of 2 (Fig. 3A). Thus, this binding isotherm suggests the involvement of two sites for Cd 2ϩ binding. Cd 2ϩ titrations in the presence of 0.25 mM Hg 2ϩ also yielded a pure exothermic reaction but showed a midpoint of binding heat changes at 1 stoichiometric equivalent (Fig. 3B). This apparent loss of one equivalent Cd 2ϩ binding site suggests that one of the two Cd 2ϩ binding sites is blocked by Hg 2ϩ binding, whereas the other is not affected. Therefore, a binding model for two sets of independent binding sites was used to fit the corresponding competitive and noncompetitive Cd 2ϩ bindings. All parameters (n, K a , and ⌬H) were allowed to float during the least-squares minimization process, resulting in an excellent fit correspond-  Table I.
Ϫ6.1 Ϯ 0.7 kcal/mol. The binding stoichiometries for both sites are close to unity, consistent with the qualitative assessment that there are two independent binding sites for Cd 2ϩ binding. One tight (site 1) and one weaker (site 2) site can be mathematically distinguished by a 29-fold difference in their binding affinities. The Cd 2ϩ binding isotherm obtained in the presence of 0.25 mM Hg 2ϩ was fit to a one-site model, resulting in best-fit parameters (K a ϭ 0.50 Ϯ 0.05 M Ϫ1 , n ϭ 0.77 Ϯ 0.03, and ⌬H ϭ Ϫ7.6 Ϯ 0.3 kcal/mol) in agreement with those of site 2. A comparison between the binding parameters obtained from free YiiP and Hg 2ϩ -prebound YiiP suggests that site 1 is a common competitive binding site for both Cd 2ϩ and Hg 2ϩ .
Cd 2ϩ titrations in the presence of 0.5 mM Zn 2ϩ broadened the binding isotherm and caused a greatly reduced exothermic heat effect, indicative of a diminished Cd 2ϩ binding to the Zn 2ϩ -prebound YiiP (Fig. 2C). The resulting featureless isotherm precluded fits of any binding model, but it was clear that Zn 2ϩ binding obstructed Cd 2ϩ binding to both site 1 and site 2.
Calorimetric Titrations of YiiP with Hg 2ϩ -The binding of Hg 2ϩ was shown to block an exothermic Zn 2ϩ binding site (Zn 2ϩ site 1) or a high affinity Cd 2ϩ binding site (Cd 2ϩ site 1). To further examine whether the binding of Zn 2ϩ or Cd 2ϩ could mutually block the Hg 2ϩ binding, Hg 2ϩ titrations were carried out to examine Hg 2ϩ binding to free-YiiP and to YiiP pre-bound with Zn 2ϩ or Cd 2ϩ . Similar to Cd 2ϩ titrations at neutral pH, Hg 2ϩ titrations showed a pure exothermic reaction with two distinct phases (Fig. 4A). In the first phase the exothermic heat effects dropped rapidly and then relaxed in the second phase to the dilution level, which was considerably higher than when titrated with Zn 2ϩ or Cd 2ϩ . After correction for heats of Hg 2ϩ dilution, an excellent fit to the Hg 2ϩ binding isotherm was obtained for a model containing two sets of independent bind-   Table I. ing sites with all six fit parameter floating freely. The resulting K a value for the tight binding site (Hg 2ϩ site 1) differs by 550-fold from that of the weaker site (Hg 2ϩ site 2) (Table I). Furthermore, the fit of the Hg 2ϩ binding isotherm obtained from the Cd 2ϩ -prebound YiiP (Fig. 4B) showed only a single Hg 2ϩ binding site with three best-fit parameters (K a , n, and ⌬H) in agreement with those of Hg 2ϩ site 2 (Table I). This result suggests that Cd 2ϩ binding blocks Hg 2ϩ site 1 but has little effect on Hg 2ϩ site 2.
The Hg 2ϩ binding isotherm obtained from Zn 2ϩ -prebound YiiP showed that the rapidly declining phase of the exothermic Hg 2ϩ heat effect was abolished (Fig. 4C). The binding isotherm was complicated, containing at least two low affinity binding sites. Thus, we were not able to quantitatively assess the effect of Zn 2ϩ on individual Hg 2ϩ binding site. However, Zn 2ϩ binding was clearly shown to inhibit the high affinity Hg 2ϩ binding (Hg 2ϩ site 1).
Taken together, calorimetric titrations revealed at least two binding sites for each of Zn 2ϩ , Cd 2ϩ , and Hg 2ϩ . All three metal ions bind to a common site (site 1) in a mutually competitive fashion. This common binding site is thermodynamically distinguishable as the exothermic binding site for Zn 2ϩ and the higher binding affinity sites for Cd 2ϩ and Hg 2ϩ , respectively. The relationship between the second sites is less clear. The Cd 2ϩ site 2 and Hg 2ϩ site 2 are independent of each other, but Zn 2ϩ binding to the Zn 2ϩ second site(s) appeared to interfere with bindings to both Cd 2ϩ and Hg 2ϩ second sites.
The pH Dependence of Metal Ion Binding to YiiP-To test the involvement of histidine residues in metal ion binding, we examined the pH dependence of Zn 2ϩ , Cd 2ϩ , and Hg 2ϩ titrations. The binding of a metal ion to the imidazole group (pK a ϭ 6.2) of a histidine is coupled to the release of a proton. Acidic conditions would prevent proton release, thereby inhibiting metal binding. YiiP was purified with MES buffer at pH 5.5, and ITC experiments were performed in the identical MES mobile phase. Zn 2ϩ titrations exhibited a predominantly endothermic reaction except that the early midrange of titrations consisted of biphasic heat effects, a fast endothermic process followed by a slow exothermic process (Fig. 5A). Deconvolution of this complicated binding isotherm was difficult. However, a comparison with the Zn 2ϩ binding isotherms obtained at pH 7.0 ( Fig. 2A) revealed a conspicuous loss of the exothermic heat profile, indicating that Zn 2ϩ binding to the common site (Zn 2ϩ site 1) was significantly inhibited at pH 5.5. The heat effects generated by Cd 2ϩ titration remained purely exothermic, but the amplitude of the binding enthalpy was also significantly reduced (Fig. 5B). The best-fit of the Cd 2ϩ binding isotherm to a two-site binding model yielded K a1 ϭ 3.15 Ϯ 0.79 M Ϫ1 , n 1 ϭ 0.31 Ϯ 0.01, and ⌬H 1 ϭ Ϫ0.06 Ϯ 0.16 kcal/mol; K a2 ϭ 0.15 Ϯ 0.01 M Ϫ1 , n 2 ϭ 1.34 Ϯ 0.02, and ⌬H 2 ϭ Ϫ3.0 Ϯ 0.08 kcal/mol. In addition to the decreases in binding enthalpy that occurred at both sites, the binding stoichiometries decreased from 1.2 to 0.31 for the high affinity site but increased from 0.84 to 1.34 for the low affinity site (Table I). The major pH effect on Cd 2ϩ binding appears to occur at the high affinity site because of the significant reduction of both ⌬H 1 and n 1 . Nevertheless, a quantitative evaluation is not possible based on a simple comparison of fit parameters obtained in two pH buffers of different proton ionization enthalpies. In contrast to the strong pH dependences of Zn 2ϩ and Cd 2ϩ , lowering the pH to 5.5 had little effect on Hg 2ϩ binding. The Hg 2ϩ binding isotherm obtained at pH 5.5 (Fig. 5C) was similar to that obtained at pH 7.0 (Fig. 4A). The best-fit to two sets of binding sites yielded K a1 ϭ 380 Ϯ 90 M Ϫ1 , n 1 ϭ 0.71 Ϯ 0.01, ⌬H 1 ϭ Ϫ22.8 Ϯ 0.2 kcal/mol; K a2 ϭ 0.4 Ϯ 0.07 M Ϫ1 , n 2 ϭ 0.94 Ϯ 0.02, ⌬H 2 ϭ Ϫ9.9 Ϯ 0.3 kcal/mol. These fit parameters are in the same range as those obtained at pH 7.0 (Table I), indicating that the Hg 2ϩ binding to both site 1 and site 2 is retained at pH 5.5. The apparent lack of pH effect also suggests that the Hg 2ϩ binding is not coupled to the deprotonation of histidine residues.
Coupling of Cd 2ϩ Binding to Deprotonation-To further evaluate the pH dependence of the Cd 2ϩ binding, the apparent enthalpy change, ⌬H app , upon Cd 2ϩ binding was obtained at pH 7.0 in three different pH buffers: HEPES, MOPSO, and ACES. The enthalpy change resulting from a possible proton exchange between a Cd 2ϩ binding site and the pH buffer would contribute to the ⌬H app value. Therefore, correlating ⌬H app with the ionization enthalpy of the pH buffer would provide a quantitative evaluation of the binding-to-deprotonation coupling. Cd 2ϩ binding isotherms were deconvoluted to yield two FIG. 4. ITC analyses of Hg 2؉ binding to YiiP. A, titrations with 5-l injections of 1 mM HgCl 2 into 0.032 mM YiiP in a 1.4-ml reaction cell containing 20 mM HEPES, pH 7.0, 100 mM NaCl, 12.5% glycerol, 0.05% DDM. B, experiment identical to A except that 0.5 mM CdCl 2 was added to the reaction cell before titrations. C, experiment identical to A except that 0.5 mM ZnCl 2 was added to the reaction cell before titrations. The solid lines represent the best-fit to a binding model including two sets of independent sites in A or one set of independent sites in B. The resulting fit parameters are summarized in Table I. elementary ⌬H app values, corresponding to the binding to site 1 and site 2, respectively. The resulting ⌬H app is related to the ionization enthalpy of the pH buffer in which Cd 2ϩ titrations were performed (Fig. 6) (23). Linear regressions yielded a slope of Ϫ1.23 for site 1 and Ϫ0.15 for site 2. According to the accumulative relationship, ⌬H app ϭ ⌬H bind ϩ n⌬H ion , the slopes of these linear regressions indicate that 1.23 protons are released from Cd 2ϩ site 1, whereas 0.15 protons are released from Cd 2ϩ site 2 upon Cd 2ϩ binding at neutral pH. These results indicate that Cd 2ϩ binding to the common site is coupled to a deprotonation process.
Effect of DEPC on Metal Ion Binding-The coupling between Cd 2ϩ binding and a deprotonation process at neutral pH is consistent with the involvement of histidine residues in the binding reaction. The role of the histidine residues was further investigated by examining zinc calorimetric titrations before and after His-specific modification by DEPC (24). The control experiment in which YiiP was titrated with Zn 2ϩ before DEPC modification showed a mix of exothermic and endothermic reactions with a characteristic exothermic-to-endothermic transition occurring in the midrange of titrations (Fig. 7A). Because the exothermic reaction was assigned to Zn 2ϩ binding to the common binding site, changes in the exothermic profile were used as a thermodynamic signature to assess the effect of DEPC modification on the common binding site. A prolonged incubation with DEPC (12 h) abolished the exothermic reaction (Fig. 7B), whereas a short incubation (0.5 h) resulted in a loss of the general exothermic profile along with a broadening of the binding isotherm (Fig. 7C). To examine the specificity of the inhibitory effect of DEPC, DEPC modifications were carried out in the presence of 2 mM Zn 2ϩ or Cd 2ϩ . Both Zn 2ϩ and Cd 2ϩ binding prevented the loss of the exothermic profile to varied degrees after 0.5 h of DEPC incubation (Fig. 7, D and E). Taken together, the DEPC inhibitory effect and the binding protection suggest the involvement of a histidine residue(s) in the common binding site. Furthermore, the cross-protection by Cd 2ϩ on the exothermic Zn 2ϩ binding site again indicates that the exothermic Zn 2ϩ binding site is a common binding site for both Zn 2ϩ and Cd 2ϩ . DISCUSSION The experiments described herein provide direct energetic measurements of heat flow derived from the interactions between YiiP and its physiological substrate Zn 2ϩ and two related group-12 metal ions, Cd 2ϩ and Hg 2ϩ . Deconvolution of binding isotherms indicates that the observed heat flows are the results of cumulative contributions from bindings to at least two thermodynamically distinguishable binding sites (site 1 and site 2) for each metal ion. The physical identities of these hypothetical binding sites were deduced from comparing the changes of binding isotherms in the absence and presence of a competing metal ion at a saturation concentration. The Cd 2ϩ or Hg 2ϩ titrations under mutually competing conditions yielded a reduced binding isotherm that could be fit to a onesite mode with binding parameters matching that of their respective second sites. These quantitative analyses are internally consistent with the assignments that Cd 2ϩ site 1 and Hg 2ϩ site 1 overlap a common binding site, whereas site 2 of Cd 2ϩ and Hg 2ϩ are two independent binding sites. Classification of the Zn 2ϩ binding sites was aided by the characteristic exothermic heat reaction attributed to the Zn 2ϩ binding to its site 1. This thermodynamic signature was abolished by the binding of Cd 2ϩ or Hg 2ϩ , indicating that Zn 2ϩ site 1 overlaps a binding site common to both Cd 2ϩ and Hg 2ϩ , the common competitive binding site. Binding affinities to site 1 are 3 M, 0.12 M, and 1.04 nM for Zn 2ϩ , Cd 2ϩ , and Hg 2ϩ , respectively. The relationship among the second sites of the three metal ions remains obscured. Cd 2ϩ or Hg 2ϩ titrations in the presence of a saturated concentration of Zn 2ϩ only yielded a low affinity binding isotherm that could not be fit to any binding model. Qualitative assessments suggested that Zn 2ϩ binding inhibited both site 1 and site 2 for Cd 2ϩ and Hg 2ϩ . Because Cd 2ϩ site 2 is independent of Hg 2ϩ site 2, the inhibitory effects of Zn 2ϩ on both Cd 2ϩ and Hg 2ϩ site 2 suggest that Zn 2ϩ might have additional binding sites. This possibility is also reflected by the poor fit of the Zn 2ϩ binding isotherm to a two-site model, indicative of inadequate modeling with two binding sites. However, deconvolution of multiple binding sites is not conclusive without additional fitting constraints, precluding a definite assignment for all Zn 2ϩ binding sites.
Metal binding is likely coupled to a deprotonation process involving three major ligand groups in proteins: imidazole, sulfhydryl, and carboxylate (25). If this is the case, the binding reaction would be a pH-dependent process. Because an extreme acidic or alkaline condition would denature YiiP, the pH dependence of metal binding was examined at pH 5.5 and 7.0, representing the protonation and deprotonation pH of the histidine imidazole. Lowing the pH from 7.0 to 5.5 caused a significant reduction of binding enthalpies associated with Zn 2ϩ and Cd 2ϩ binding to the common binding site, suggesting the involvement of histidine residues. This finding is consistent with the inhibitory effect of DEPC modification on the binding isotherm of Zn 2ϩ titrations, showing a specific inhibition of the exothermic reaction. The deprotonation associated with Cd 2ϩ binding was further indicated by the linear relationship between the total heat exchange and the ionization heat of the pH buffer. The slope of this linear relationship suggests that 1.23 protons are released upon Cd 2ϩ binding to the common binding site. It is not known to what extent the histidine residue in the binding site is deprotonated at pH 7.0. These 1.23 protons released upon Cd 2ϩ binding could come from the deprotonation process of two possible ligand groups: a partially protonated histidine imidazole and a cysteine thiol that is expected to be fully protonated at neutral pH. Interestingly, in a parallel titration experiment with Hg 2ϩ , the lower pH had little effect on the binding of Hg 2ϩ to both site 1 and site 2. The lack of pH dependence of the Hg 2ϩ binding to the common binding site contradicts the pH dependence of Zn 2ϩ and Cd 2ϩ binding to the same site. This apparent discrepancy is likely attributed to the difference in coordination chemistry. Zn 2ϩ and Cd 2ϩ binding generally takes up a 4-or 5-coordinate geometry, whereas the binding of Hg 2ϩ prefers a lower coordination number of 2 or 3 (30). The lower coordination number of Hg 2ϩ binding may allow of a high Hg 2ϩ binding affinity without the need to interact with the histidine that is essential to Zn 2ϩ and Cd 2ϩ coordination. In this case, a sulfhydryl group is another likely ligand group in the common site in addition to the imidazole of a histidine, because sulfur is a preferred donor atom for Hg 2ϩ coordination in a linear or trigonal geometry (31). The identification of determinant residues in the common binding site is under way by site-specific labeling of histidine and cysteine residues.
Dissection of the Gibbs free energy of binding into entropic and enthalpic changes provides thermodynamic details of the molecular interactions between YiiP and metal ions. The binding enthalpy change primarily reflects the strength of metal ion-YiiP interactions relative to that of metal ion-solvent interactions, whereas the binding entropy change mainly depends on the gain of solvation entropy because of metal ion desolvation relative to the loss of protein conformational entropy (freedom) upon metal ion binding. The binding enthalpy change at the common binding site is Ϫ4.5 kcal/mol for Zn 2ϩ , becomes slightly more favorable at Ϫ6.5 kcal/mol for Cd 2ϩ , but leaps to Ϫ22.5 kcal/mol for Hg 2ϩ . The corresponding T⌬S values are 3.0 kcal/mol for both Zn 2ϩ and Cd 2ϩ binding and a negative value of Ϫ10.3 kcal/mol for Hg 2ϩ binding to the same site. Therefore, the binding of Zn 2ϩ and Cd 2ϩ to the common site are driven both enthalpically and entropically, whereas the binding of Hg 2ϩ is driven only by enthalpy. The unfavorable entropy change during Hg 2ϩ binding is dominated by a much more pronounced, favorable enthalpy change. The modest thermodynamic difference between Zn 2ϩ and Cd 2ϩ binding implies that Cd 2ϩ and Zn 2ϩ bind to the common binding site through liganding to a set of mostly shared donor groups with similar, if not identical, coordination geometry. On the other hand, Hg 2ϩ appears to adopt a different thermodynamic approach to reach a much higher binding affinity. The distinctive thermodynamic character of Hg 2ϩ binding suggests a much stronger bond formation and a greater degree of structural rigidification as compared with Zn 2ϩ and Cd 2ϩ binding. These differences, corroborated by the pH independence of the Hg 2ϩ binding to the common binding site, suggest that the binding of Hg 2ϩ to the common binding site may involve a different set of donor groups arranged in a different coordination geometry.
A comparison of the binding of three metal ions to their respective second sites indicates that the binding of Cd 2ϩ and Hg 2ϩ is favored both enthalpically and entropically, whereas the binding of Zn 2ϩ to its site 2 is associated with an unfavorable enthalpy change, leaving the entropic force to drive the FIG. 7. Inhibitory effect of DEPC on Zn 2؉ binding to YiiP. All titrations were preformed at 25°C with 5-l injections of 1.5 mM ZnCl 2 into 0.09 mM YiiP in 20 mM HEPES, pH 7.0, 100 mM NaCl, 12.5% glycerol, 0.05% DDM, 0.25 mM TCEP. A, YiiP without DEPC modification. B, YiiP was incubated with 10 mM DEPC for 12 h. The resultant DEPC-modified YiiP was purified by HPLC before calorimetric titrations. C, YiiP was incubated with 10 mM DEPC for 30 min and then HPLC-purified. D, experiment identical to C except that 2 mM ZnCl 2 was added to YiiP before DEPC treatment. E, experiment identical to C except that 2 mM CdCl 2 was added to YiiP before DEPC treatment.
binding event (Table I). The binding enthalpy changes are Ϫ6.1 and Ϫ4.1 kcal/mol for Cd 2ϩ and Hg 2ϩ respectively, but a positive value of 12.5 kcal/mol for Zn 2ϩ . The 16.5-18.5 kcal/mol unfavorable ⌬⌬H for Zn 2ϩ binding is overcome by a significant gain of T⌬S (17.1 kcal/mol) to give rise to a Gibb free energy of binding of Ϫ5.2 kcal/mol, not too far away from that of Cd 2ϩ (Ϫ7.4 kcal/mol) and Hg 2ϩ (Ϫ8.5 kcal/mol). The distinctive Zn 2ϩ enthalpic and entropic contributions to the binding free energy reflect a weaker coordination environment and an increased degree of disorder and desolvation upon Zn 2ϩ binding to its site 2. This entropically driven Zn 2ϩ binding suggests that Zn 2ϩ site 2 may be thermodynamically distinct from that of Cd 2ϩ and Hg 2ϩ , which are, by virtue of the lack of mutual binding competition, independent of each other. It is not clear whether the inhibitory effect of Zn 2ϩ on both Cd 2ϩ site 2 and Hg 2ϩ site 2 is due to a physical overlapping of binding sites or an allosteric effect resulting from the increased disorder induced by Zn 2ϩ binding.
Metal ion binding to a CDF protein represents the first step of the transport process. A stopped-flow analysis of the homologous E. coli CDF protein ZitB suggested that the CDF transport process is consistent with the following two-step scheme, M ϩ T 1 L | ; where T 1 and T 2 correspond to the inward facing and outward facing conformations of the transporter, and M is the metal ion (16). The kinetic step MT 1 3 M ϩ T 2 represents the translocation of a metal ion across the membrane. The relationship among three rate constants k 1 , k Ϫ1 , and k 2 is defined by K m ϭ (k 2 ϩ k Ϫ1 )/k 1 . K m for both Zn 2ϩ and Cd 2ϩ are in the order of 100 M, and the transport turnover rate k 2 is in the order of 5 s Ϫ1 (16). Kinetic studies of YiiP showed that Zn 2ϩ and Cd 2ϩ are both effective transport substrates with kinetic parameters in the same order of magnitude of the ZitB parameters. 3 In this study, calorimetric titrations of YiiP with Zn 2ϩ and Cd 2ϩ revealed multiple binding sites with binding affinities in the order of the micromolar to submicromolar range. It is not known which one of these binding sites is involved in the transport process and the binding affinity of which conformational states corresponds to the values determined by calorimetric titrations. Nevertheless, our thermodynamic analyses provide an estimate of the transport K d (k Ϫ1 /k 1 ). If it is assumed K d ϭ 1 M, K m ϭ 100 M and k 2 ϭ 5 s Ϫ1 , the corresponding on and off rates of the metal ion-CDF complex can be calculated as follows: k 1 ϭ 50,000 s Ϫ1 M Ϫ1 and k Ϫ1 ϭ 0.05 s Ϫ1 . k Ϫ1 is slower than k 2 by a factor of 100, indicating that the metal-protein complex (MT 1 ) will undergo a MT 1 3 T 2 ϩ M conformational transition with a likelihood of being 100-fold greater than that of releasing it back to the cytosol (MT 1 3 T 1 ϩ M). Thus, the transport process is nearly irreversible once a metal ion is loaded to the binding site. However, the binding of metal ion appears to be a highly unfavorable kinetic process under the physiological condition. It has been estimated that intracellular zinc binding capacity exceeds the total number of zinc ions in a cell. This overcapacity sets the biologically available zinc pool at an extraordinarily low level (26). The zinc concentration in the cytoplasmic milieu is estimated to be less than 10 Ϫ10 M (27). The rate of zinc binding at this concentration is 5 ϫ 10 Ϫ6 s Ϫ1 (equals to k 1 ϫ 10 Ϫ10 ), giving a half-time of Zn 2ϩ loading to a CDF in the order of 40 h. Therefore, the kinetic barrier of zinc binding is quite prohibitive under zinc homeostatic conditions. The function of CDF transporters may not be driven by Zn 2ϩ thermodynamic equilibrium but, rather, operated in a kinetically controlled substitution fashion. Several recent studies have suggested that a class of metallochaperone proteins might be responsible for catalyzing metal ion transfers from protein donors to acceptors with high metal binding affinity, while retaining a fast transfer kinetics (21,28,29). This traffic of metal ions in the cell allows the delivery of metal ions through specific protein-protein interactions to overcome the kinetic binding barrier at an extremely low free metal-ion concentration. Likewise, the efflux or sequestration of cytosolic zinc through CDF transporters may require a cytoplasmic zinc trafficking factor that can capture cytoplasmic zinc and deliver it to the binding site of a CDF to initiate the transport process.